It is desirable, in a variety of contexts, to capture signals. The nature of the signals may vary from context to context. For example, in a medical context, it may be desirable to capture signals that represent heart activity. In the context of sound studios, it may be desirable to capture audio signals produced by music artists. The techniques described herein are not limited to any particular type of recorded signal. The digital representation of a signal is referred to herein as “signal data”.
For a variety of reasons, it may be desirable to edit the signal data after a signal has been digitally recorded. Many signal editing applications are available for performing post-recording edits to a captured signal. In the context of audio signals, many audio editing applications allow a user to listen to the audio. While the audio is being played, the user is presented with a visual representation of the signal, with an indication of the location, within the signal, that is currently being played. While listening to the audio and watching the visual representation of the signal, the user may identify a problem that requires fixing. For example, the user may hear a “cough”, and see a spike that represents the cough in the visual representation of the signal. The user may then use a tool provided by the editing application to correct the problem. For example, the user may replace the portion of the signal that contains the cough with an ambient noise print, as described in U.S. patent application Ser. No. 11/104,995, filed on Apr. 12, 2005, the contents of which are incorporated herein by this reference.
Sometimes, it is desirable change the length of a recorded signal relative to a particular domain, such as time. For example, in the context of audio signals, it may be desirable to convert a 17 minute lecture into a 15 minute lecture. Some digital signal editing applications may provide mechanisms for accomplishing such wholesale conversions. However, such conversions may not produce optimal results. For example, there may be a portion of the 17 minute lecture in which the speaker talks so quickly that it is difficult to catch all of the words. If the duration of the entire lecture is reduced, then all speech within the lecture will be played back at a rate that is even faster than the original rate. Consequently, after the conversion, the portion of the lecture in which the speaker talks quickly will be even more difficult to understand.